Data anomaly detection method for cigarette real-time data acquisition

Abstract

The invention discloses a data anomaly detection method for cigarette real-time data acquisition. The design concept of the method is as follows: a machine learning algorithm is applied, historical data characteristic value prediction conditions are fully combined, and total data are subjected to full sample analysis, so that high-precision detection is automatically carried out on the anomaly condition of the real-time data acquisition data. The method effectively solves the problem that at present, the pain point of overall data is estimated only through current sample data, and can perform systematic analysis on the authenticity, effectiveness and accuracy of the data so as to guarantee that the data quality is real and controllable.

Description

technical field [0001] The invention relates to the field of tobacco industry, in particular to a data anomaly detection method for real-time data collection of cigarettes. Background technique [0002] At present, cigarette companies mainly rely on the Wincc Server server of the centralized control system when collecting real-time data on the spot, especially for the bottom sensor data, and transmit it to the cloud through Profinet and Profibus-DP. In the process, the detection of data quality is particularly important. [0003] However, in the data collection process of the real tobacco industry site, data quality problems often occur frequently. According to the analysis, this is mainly caused by the following reasons: First, the data source problem, the data source itself has noise data, and the data has The situation of manual operation may lead to inconsistencies in the overall situation of the data; the second is the problem of data extraction time point. The product...

AI Technical Summary

Problems solved by technology

[0005] However, the problem with this traditional method is that the idea of ​​using samples to test the population often falls into two types of errors, α and β, which cannot be avoided in mathematical statistics, that is, the sample may be better than the population or the sample may be inferior to the population, resulting in distortion of data quality

Images